Open-source agentware that joins video calls with full memory of your work — ready to discuss, present, and act. Local-first.
OpenClaw
Claude Code
CallingClaw is agentware — it doesn't replace your agent, it gives your agent a meeting room, a calendar, and a voice. Your agent's capabilities come along naturally.
Point at the screen and say what to change — your agent sees exactly where and what. No need to deploy to the public web for review — everything stays local.
The full meeting cycle, agentic — from prep to delivery.
Download the desktop app. A guided setup walks you through each permission — screen, mic, accessibility, then connects to your agent.
Message your agent on Telegram or CLI. It schedules meetings, prepares context from your recent work, and stays ready.
Your agent joins Google Meet, shares screen with browser tabs, moves a visible cursor, and speaks with live captions.
| Task | Status | Priority | Assignee |
|---|---|---|---|
| Implement voice routing | Done | P0 | |
| Add screen share API | In Progress | P0 | |
| Calendar integration | Blocked | P1 | Peter |
| Visual diff optimization | Done | P1 |
Every action item is paired with a screenshot and timestamp. Your agent already knows what was discussed — alignment is immediate.
| Screenshot | Discussion & Action |
|---|---|
|
|
23:16
Hero Copy Update
Emphasize agent comes with full memory, not blank. Update tagline and subtitle.
Key
|
|
|
23:22
Footer Layout Bug
Column alignment broken on mobile. Fix spacing after copy update ships.
Info
|
Traditional meeting AI records and transcribes. CallingClaw joins, speaks, and acts — locally, with full access to your machine.
| CallingClaw | Pika Stream | Granola / Otter | |
|---|---|---|---|
| Real-time voice participation | Full duplex | Partial | — |
| Screen perception & sharing | Vision + DOM | — | — |
| Computer control | Voice-triggered | — | — |
| Agent MCP action | Full tool use | Limited | — |
| Meeting preparation | Automatic | — | — |
| Cross-meeting memory | Local file memory | Limited | Transcript only |
| Live action item capture | Real-time | — | Post-meeting |
| Data privacy | 100% local | Cloud | Cloud |
Like Thinking, Fast and Slow — fast voice for real-time dialogue, slow reasoning for deep execution, connected through shared memory.
Conversational, intuitive, immediate. Handles real-time dialogue, answering questions, guiding discussions, and aligning context through natural voice.
Analytical, methodical, thorough. Reads files, accesses memory, plans multi-step execution. Prepares meeting briefs before, executes action items after.
Apache 2.0 license. Full codebase on GitHub.
No subscription, no fees. Bring your own API keys.
All agent data stays on your machine.
Download the desktop agentware. GUI walks you through setup and authorization, then your agent runs via CLI in the background.